J. Agric. Food Chem. 1997, 45, 850−858
850
Determination and Origin of the Aroma Impact Compounds of Yogurt Flavor Andreas Ott, Laurent B. Fay, and Alain Chaintreau* Nestle´ Research Centre, Nestec Ltd., P.O. Box 44, Vers-chez-les-Blanc, CH-1000 Lausanne 26, Switzerland
Yogurt exhibits a delicate and low intense flavor which requires mild sample isolation techniques and sensitive identification means. This paper decribes the application of two new flavor separation techniques: combined static and dynamic headspace, and preparative simultaneous distillationextraction under vacuum. Of the 91 components identified by mass spectrometry, 21 were found to have a major impact on the aroma using a new method for the treatment of GC-sniffing data. One of these compounds, 1-nonen-3-one, was identified by spectral means for the first time in a flavor. Its odor detection threshold of 8 pg/kg appears to be one of the lowest among known flavor compounds. On the basis of the peak areas of yogurt and milk aromagrams, some compounds seem to be already present in milk, others appear to be generated by the strains. Keywords: Yogurt aroma; headspace; GC-olfactometry; impact-compound; composition INTRODUCTION
Since the 1950s, yogurt volatiles have given rise to many publications (for review, see Maarse, 1994), and more than 60 flavor compounds have been identified so far. Its consumption has become very popular since the 1960s due to the delicate flavor associated with its particular texture and the claimed health benefits of lactic acid bacteria for humans. Headspace and simultaneous distillation-extraction (SDE) have been commonly used for the analysis of its flavor. Because of the low intensity of yogurt odor, the sample is often heated to increase the volatility of the flavorings (Go¨rner et al., 1968; Hild, 1979; Ulberth, 1991; Kang et al., 1988). This can alter the composition of this sensitive aroma, and a mild sampling technique must be applied. Using a purge-and-trap sampling at a lower temperature, Laye and Imhof respectively identified 23 and 33 compounds (Laye at al., 1993; Imhof and Bosset, 1994; Imhof et al., 1994a,b). Among 19 papers listed in the TNO compilation, none has tried to determine the key volatiles responsible for the typical yogurt aroma. Recently, Imhof suggested 2,3-butanedione, 2,3-pentanedione, dimethyl sulfide, and benzaldehyde to be the most potent ingredients (Imhof et al., 1995). However, this selection was indirect, as it was based on odor unit values deduced from the author’s quantitations and from literature thresholds. Hyphenated gas chromatography-olfactometric techniques allow the direct determination of impact odorants in foods. Most commonly applied methods are Charm analysis (Acree et al., 1984) and aroma extract dilution analysis (AEDA) (Ullrich and Grosch, 1987). Both methods are based on GC-olfactometry of an aroma extract which is diluted until no odor is detected at the sniffing port. More recently Grosch’s group sampled the headspace surrounding the food: its composition better represents the smell that is perceived by the consumer (Semmelroch and Grosch, 1995). Since the vapor phase odor is first perceived when opening a yogurt pot, its quality will influence the consumer’s preference. Because the retronasal percep* Corresponding author (e-mail Alain.Chaintreau@ chlsnr.nestrd.ch). S0021-8561(96)00508-0 CCC: $14.00
tion is also sensitive to the headspace odor released by the yogurt, we decided to focus on the headspace and to determine the main odor contributors by direct measurement of their individual impact. As critical evaluation of dilution sniffing methods pointed to reproducibility problems (Abbott et al., 1993), we recently developed another approach. Only one dilution level was used, but GC-olfactometry was repeated several times. Aromagrams of individuals were then “averaged” to get rid of previously mentioned reproducibility problems. Resulting aromagrams allow odor profile comparisons as peak intensities are related to the frequencies of odor detection: “NIF” and “SNIF” (respectively “Nasal Impact Frequency” and “Surface of Nasal Impact Frequency; Ott et al., 1996). A detailed paper will report this headspace/GC/sniffing method (Pollien et al., in press). MATERIALS AND METHODS Yogurt Preparation. Pasteurized full fat milk from a local retailer (Orlait SA, Lausanne, Switzerland) was fortified with 2.5% of skimmed milk powder and heat treated at 98 °C for 15 min, immediately cooled to about 35 °C in a water bath and inoculated with a starter culture of Streptococcus thermophilus and Lactobacillus delbrueckii ssp. bulgaricus. The prepared solution was filled in sterile pots of 150 mL, hermetically sealed, and fermented at 41 °C during 5 h (until pH 4.2 was reached). Immediately after termination of fermentation the yogurt samples were cooled to 4 °C and stored during 2 weeks at this temperature prior to analysis. Headspace Sampling. Dynamic Headspace/GC/MS (Procedure 1a, Used for MS Identification). A 10 g yogurt sample was dispersed into 5 mL of a saturated salt solution and filled into the sparger of a purge-and-trap system (Tekmar Co., Cincinnati, OH). Samples were purged with ultrapure nitrogen at a flow rate of 40-60 mL/min during a time period varying from 5 to 99 min. Volatiles were thermally desorbed from Tenax (180 °C, 4 min), cryofocused, and injected into an HP 5995 GC/MS (Hewlett-Packard, Avondale, PA) equipped with a FFAP (Supelco Inc., Bellefonte, PA) column, 30 m length, 0.25 mm i.d., 0.25 µm phase thickness. Helium was used as carrier gas at a 0.65 mL/min flow. The column was kept at 20 °C for 1.0 min, increased at a 4 °C/min rate from 20 to 200 °C, and maintained for 1 min at 200 °C. The mass spectra were acquired with a source temperature of 220 °C, under a 70 eV ionization potential. Mass spectra and retention indices of the compounds detected were compared with those in our user-generated libraries.
© 1997 American Chemical Society
J. Agric. Food Chem., Vol. 45, No. 3, 1997 851
Aroma Impact Compounds of Yogurt Flavor Linear retention indices were determined (Van den Dool and Kratz, 1963) after injection of a n-alkane series (C5-C17), under the same conditions. Headspace Traps/GC/MS (Procedure 1b, Used for MS Identification). 100 g of yogurt and 16 g of distilled water were poured into a 3-neck vessel of 500 mL volume, thermostatted to 30 °C, and rapidly stirred with a magnetic bar. Volatiles were purged with an ultrapure helium flow of 30 mL/min and trapped on Tenax tubes for 60 min. Volatiles were thermally desorbed from Tenax (250 °C, 5 min), using a TD-2 thermal desorber (Scientific Instruments Services Inc., Ringoes, NJ) at 250 °C. They were cryofocused and injected into an HP 5890 GC (Hewlett-Packard) equipped with a DB-Wax column (J&W Scientific, Folsom, CA), (30 m length, 0.25 mm i.d., 0.25 µm phase thickness). Helium was used as carrier gas at a 0.65 mL/min flow. The column was kept at 20 °C for 5 min, increased at a 4 °C/min rate to 200 °C, and maintained for 10 min. The column was directly coupled with the EI source of a HP 5971 mass spectrometer operating in the same conditions as mentioned above, in scan or in SIM mode. Criteria for identification of compounds were the same as previously mentioned. Headspace Cell/GC/Sniffing or FID (Procedure 1c; Chaintreau et al., 1995). This procedure was used for all sniffing experiments and to test the repeatability of the headspace/ GC method. An aliquot of 25 g of yogurt was placed into the sample space of the headspace cell and equilibrated during 2 h at 30 °C in a water bath. The headspace of the cell (160 mL) was then passed through the trap containing 250 mg of Tenax with a flow of 40 mL/min. Volatiles were thermally desorbed from Tenax (300 °C, 15 min), using an ATD400 thermal desorber (Perkin-Elmer Corp., Norwalk, CT). They were refocused on a internal cold Tenax trap (-30 °C) and desorbed at 300 °C for 3 min into an HP 5890 GC (Hewlett-Packard) equipped with a DB-Wax column (J&W Scientific) (60 m length, 0.53 mm i.d., 1.00 µm phase thickness). Helium was used as carrier gas at a 10.6 mL/min flow. The column was kept at 20 °C for 5 min, increased at a 4 °C/min rate to 200 °C, and maintained for 10 min. The column outlet was connected either to a FID or to a sniffing port (Brechbu¨hler SA, Plan-les-Ouates, Switzerland) equipped with a humidified air make-up. After each sampling, the cell was cleaned in a vacuum oven at 50 °C under 100 mbar for at least 1 h. Tenax sampling tubes were cleaned before use by heating for 1 h to 300 °C under a helium flow (50 mL/min). Vacuum Transfer (Procedure 2; Bemelmans, 1979). 100 g of lyophilized yogurt was extracted in a Soxhlet apparatus during 30 h with 700 mL of freshly distilled diethyl ether. The extract was separated from fat by high-vacuum transfer under 10-3 mbar and trapped at liquid nitrogen temperature. Condensates were concentrated to 3 mL on a Vigreux column. The concentrate was extracted 3 times with 1 mL of a 0.5 M Na2CO3 solution and washed 3 times with 1 mL of saturated NaCl solution. The ether extract was dried over Na2SO4 and concentrated with a microdistilllation device (Bemelmans) to 100 mL. The aqueous fractions were acidified with 1 N HCl and extracted 3 times with 10 mL of distilled diethyl ether. The etheral extract was washed with 3 × 10 mL of saturated NaCl solution and dried over Na2SO4. The extract was concentrated to 100 µL with a microdistilllation device. Vacuum SDE (Procedure 3 for MS Identification of the Most Diluted Impact Components). Simultaneous distillation extraction under vacuum was used as this technique dramatically reduces the formation of thermal artefacts (Maignal et al., 1992). 2 kg of yogurt was dispersed into 2 L of previously degassed water and saturated with salt. The dispersion was SDE extracted during 7 h in a preparative apparatus under 40 mbar, using continuous sample feeding according to a reported procedure (Chaintreau et al., 1995). Freshly distilled butyl ethyl ether (5 mL) (Fluka AG, Buchs Switzerland) was used as the solvent. Operating temperatures were 39, 15, and 9.5 °C, in the sample flask, in the sample reservoir, and in the solvent flask, respectively. The resulting organic extract (ca. 1 mL) was directly used for GC-MS without prior concentration.
Gas Chromatography-Tandem Mass Spectrometry (GC/MS/MS). Mass spectra were acquired with a Finnigan MAT TSQ-700 mass spectrometer (San Jose, CA) interfaced with an HP-5890 gas chromatograph. The column was a DBWax column (J&W Scientific), 30 m length, 0.32 mm i.d., 0.25 µm phase thickness. Helium was used as carrier gas at a pressure of 10 psi. The column was kept at 60 °C for 1 min, increased at a 15 °C/min rate to 240 °C, and maintained for 2 min. Positive or negative chemical ionization with ammonia as reagent gas were used to generate [M + NH4]+ ions or [M - H]- ions, respectively, analysed by collision-induced dissociation experiments. A collision energy of 6 eV in the laboratory frame was used, with argon as collision gas set to 1 mTorr. Sniffing Procedure (Ott, 1996). Eight people experienced in GC-olfactometry were selected. Sniffing of the chromatogram was divided into two parts of about 25 min. Each person participated in the sniffing of both parts of the chromatogram, but during two distinct sessions to be alert. Eluting aroma relevant flavourings were recorded by pressing a button during the whole sensory impression. The square signal was recorded by an HP Pascal workstation. In addition, odor descriptors were registered on tape for every peak. The eight individual aromagrams of a given sample were summed to one chromatogram with home-made software, yielding the averaged aromagram. Linear retention indices were calculated as mentioned. Compounds identified by GC/ MS were reinjected into the GC/sniffing system to check their retention indices. Their odor quality was compared when necessary. Synthesis of 1-Nonen-3-one. 1-Nonen-3-one was obtained by oxidation of the corresponding alcohol by pyridinium chlorochromate (Corey and Suggs, 1975) and purified by preparative HPLC. MS and 1H and 13C NMR spectra were in agreement with published data (Nakahira et al., 1992). Odor Threshold. The nasal perception threshold of 1-nonen-3-one was measured using a triangle presentation. A series of three samples in weighing jars of 65 mL volume containing 35 mL of liquid was presented to a group of 21 trained panelists. One sample contained the stimulus dissolved in mineral water (Vittel), the other two samples were mineral water only. For each session, five different groups with increasing concentrations of the stimulus were presented and the assessors had to indicate which sample was different from the two others. All trials were performed at room temperature (21 °C). The glass vessels were washed before use with a 2% aqueous solution of RBS (Fluka AG) then rinsed with deionized and afterward with distilled water. All glass material was stored in a vacuum oven at 50 °C and 100 mbar at least overnight prior to use. The equation expressing the probit as a function of the concentration logarithm was adjusted to the experimental data using the curve-fitting software “TableCurve 2D” for Windows (Jandel Scientific, Erkrath, Germany). DISCUSSION
Headspace Sampling. Adsorption of volatiles on the glass wall of a syringe has been shown to occur (Buttery et al., 1969). Therefore we preferred the vapor phase sampling, which has a proven reliability. We recently developed a new headspace cell for the determination of partition coefficients (Chaintreau et al., 1995). Its repeatability was established with model mixtures, and the system exhibited good accuracy when applied to quantitation in the vapor phase. This technique mimics well the aroma release conditions into a yogurt pot since the flavor is allowed to equilibrate between the matrix and the headspace at a controlled temperature, under atmospheric pressure. The air surrounding the yogurt is then collected without disturbing the partition of flavorings between both phases.
852 J. Agric. Food Chem., Vol. 45, No. 3, 1997
Ott et al.
Table 1. GC/MS Identification of Yogurt Flavorings and Repeatability of GC Area retention index compounds
Tekmarb
acetaldehyde dimethyl sulfide methylcyclohexane propanal 2-propanone furan methyl acetate 2-methylfurane butanal ethyl acetate 2-butanone methanol 3-methylbutanal dichloromethane benzene 2-propanol ethanol 2-pentanone 2,3-butanedione acetonitrile chloroform toluene 2-butanol S-methyl thioacetate 1-propyl alcohol 2,3-pentanedione dimethyl disulfide butyl acetate hexanal 2-hexanone 2-methylthiophene 2-methyl-1-propanyl alcohol ethylbenzene 1,3-dimethylbenzene (1,4?) 3-penten-2-one 1,3-dimethylbenzene (1,3?) 1-methylpyrrole 3-heptanone 1-butanol 1-ethyl-4-methylbenzene 1-penten-3-ol limonene 1,4-dimethylbenzene (1,2?) 2-heptanone propylbenzene 3-methyl-2-butenal 2-pentylfuran pyrazine ethenylbenzene 1-pentanol 3-octanone 2-methyl tetrahydrofuran-3-one trimethylbenzene or ethylmethylbenzene methylpyrazine octanal 3-hydroxy-2-butanone 1-methyl ethenylbenzene 3-methyl-2-butenol 2-nonanone 2-butanol 2-hydroxy-3-pentanone dimethyl trisulfide furfural acetic acid 1H-pyrrole propionic acid benzaldehyde 2-methyltetrahydrothiophen-3-one 2-methylpropanoic acid 2-undecanone butyric acid 2-furanmethanol 3-methylbutanoic acid (phenylacetaldehyde)j pentanoic acid
709 757 756 800 802 802 834 871 877 894 908 916 919 936 939 950 956 988 993 1012 1028 1044 1049 1057 1064 1076 1081 1086 1094 1094 1102 1121 1133 1142 1148 1150 1155 1171 1179 1181 1192 1195 1200 1208 1226 1230 1252 1253 1285 1292 1300 1304 1308 1312 1317 1333 1358 1362
vacuum
transferc
yogurt
milk
area
rel. st. dev.
(e-g) (e-g) (e-g) (e-g) (e,g) (e-g) (e) (h) (f,g) (e-g) (e,g) (e-g) (h) (h) (f,g) (g) (e-g) (e-g) (e,g) (e) (e) (g) (f,g) (f) (e-g) (e) (e,g) -
331354 1862 123990 12264 8628 10505 2228 15628 2092 1912 2237 -
5.8 3.8 1 1 27.5 10.1 1.9 3.2 11.1 6.5 4.2 -
1623 1659 1667
(g) (e) (e,f) (e,f) (e) (e) (e) (e,g) (e) (e) (e) (e)
(e) (e) (e) (e) (e,i) (e) (e) (e,i) (e) (g) (e) (e,h) (e) (e) (e,i) (e,i) (e,i) (h) (e,h) (e,i) (e) (g) (e) (e) (e,i) (e) (e) (e) (e,i) (i) (i) (i) (e) (e) (g) (i) (e,i) (i) (e) (e) (e) (i) (e) (e) (e) (e,g) (e) (e) (e,i) (e,i) (e,i) (e) (e,i) (e,i) (e) (e,i) (e)
1736
(e)
(e)
1950 4541 2809 2605 -
18.1 18.7 19.5 6.8 -
1137
1182
1286 1389
1397 1396 1405 1495 1498 1547
GC areaa
literature SDEd
1442 1536
1556 1561 1567 1598
J. Agric. Food Chem., Vol. 45, No. 3, 1997 853
Aroma Impact Compounds of Yogurt Flavor Table 1 (Continued) retention index compounds 2-dodecanone hexanoic acid benzothiazole heptanoic acid 2-pentadecanone octanoic acid nonanoic acid decanoic acid γ-dodecalactone δ-dodecalactone benzoic acid
Tekmarb
vacuum
transferc
1845
2059 2164 2271 2420
GC areaa
literature SDEd
yogurt
milk
area
rel. st. dev.
1809 1845 1947 1954 2021 2060 2164 2275 2368 2420 2436
(e) (e,f) (e) (e) (e)
(e,i) (e) (e,i) (i) (i) (e,i) (i) (e,i) (e,i) (e,i) (e)
4407 -
1.2 -
a
Procedure 1c with an FID detector. b Tekmar purge and trap injection, retention indices on FFAP phase (sampling procedure 1a or b). c Direct solvent extraction and vacuum transfer, retention indices on DB-Wax phase (sampling procedure 2). d Vacuum SDE, retention indices on DB-Wax phase (sampling procedure 3). e Maarse, 1994. f Laye et al., 1993. g Imhof et al., 1995. h Imhof et al., 1994a,b. i Shiratsuchi et al., 1994. j No peak when injected using Tekmar but found using procedure 2 (ATD400/GC/FID).
Figure 1. NIF profile of the yogurt aroma isolated according to procedure 1c (retention indices at the peak apex).
As the cell repeatability was tested with model mixtures in water (Chaintreau et al., 1995), it was evaluated again using yogurt flavorings themselves, according to procedure 1c, with FID detection. Starting from the same batch, five replications of the headspace trapping were performed. For 16 GC areas reported in Table 1 relative standard deviation was less than 10% for 10 peaks, and less than 20% for 15 peaks. Methanol area variations might be attributed to possible breakthrough due to its lower absorption in Tenax than other compounds (Maier and Fieber, 1988). Further flavor generation by the strain during the analysis time probably did not occur because it would have given rise to a continuous increase or decrease of the peak abundances. As variations occurred randomly, other causes must be found (complex matrix, possible heterogeneity, etc.). However 10% or 20% was considered to be below variations recognized by the human nose with such a method (Pollien et al., in press). Flavoring Identification. Yogurts were analyzed 2 weeks after their preparation since the flavor generation is known to mainly occur during the fermentation
step and it does not significantly change during the refrigerated storage at 4 °C (Imhof and Bosset, 1994). Flavorings were isolated from their matrix using three different mild sampling methods, described in Materials and Methods as procedures 1a, 2, and 3. Table 1 summarizes results of the three isolations: most of the 86 compounds identified have already been mentioned in yogurt, milk, or skimmed milk powder, which are ingredients of the yogurt recipe. Among the newly identified compounds, some of them could contribute to the overall aroma: 1-penten-3-ol, 3-octanone, 2-methyl tetrahydrofuran-3-one, 3-methyl-2-butenol, 2-methyltetrahydrothiophene-3-one, and 2-methylpropanoic acid. Their contribution was determined by GC/ sniffing. Dichloromethane, chloroform, and acetonitrile were found: as no halogenated solvent nor acetonitrile was used in the sampling procedure 1 and since blank analyses using the Tekmar system did not exhibit any trace of these components, they seem to originate from milk itself (Maarse, 1994; Imhof and Bosset, 1994; Imhof et al., 1994a,b).
854 J. Agric. Food Chem., Vol. 45, No. 3, 1997
Ott et al.
Table 2. Odor Descriptors of Yogurt Impact Flavorings (NIF > 12.5%) ret. ind.a
a
ret. ind. (ref.)
descriptors
716 757 995 1082 1120 1221 1322 1406 1424 1462 1479
710 754 993 1082 1109 1222 1324 1403 1425 1460 1479
1551
1557
fresh, green, pungent milk, lactone-like, sulfury, warm butter, diacetyl, vanilla butter, vanilla, mild gasoline, plastic, styrene metallic, aldehydic, herbaceous mushroom, earthy sulfury, H2S, fecal mushroom, earthy pungent, acidic, vinegar soup, cooked vegetable, pungent, sulfury green, leather, sulfury
1680 1684 1715 1750 1882 1896 2002 2043
1678 1688 no data no data 1862-1885 1894 2000 no data
flowery sweaty, cheese, soy sauce, flowery flowery, warm, caramel metallic rancid, flowery bacon, phenolic, smoked, spicy burnt, rubbery hydrocarbon, chemical, burnt rubber
identification and descriptors proposed by the data-bank acetaldehyde: fresh, fruity, pungent, ether dimethyl sulfide: sulfury, roasted 2,3-butanedione: buttery, fatty, oily, vanilla 2,3-pentanedione: buttery, caramellic 2-methylthiophene: gasoline, onion-like, paraffinic 3-methyl-2-butenal: aldehydic, caprylic, pungent 1-octen-3-one: mushroom dimethyl trisulfide: strong sulfury, cabbage, alliaceous no data acetic acid: sour, vinegar-like methional: aldehydic, caprylic, cooked potato-like, green, sulfurous (2E)-nonenal + 2-methyltetrahydrothiophen-3-one: fatty, green, pungent, leather 2-phenylacetaldehyde: floral, honey-like 3-methylbutyric acid: acidic, cheese-like no data no data caproic acid: acidic, caprylic, fatty guaiacol: aromatic, burnt, phenolic, spicy benzothiazole: roasted, rubbery no data
Retention index on DB-Wax phase, using headspace injection (procedure 1c).
Table 3. Compounds Identified Using Sampling Procedures or Identification Means Other Than Those of Table 1 compound
ret. indexa
samplingb
mass spectrometric experiment
literature
1-octen-3-one 1-nonen-3-one methional 2-methyltetrahydrothiophen-3-one (2E)-nonenal phenylacetaldehyde guaiacol
1322 1424 1479 1551 1551 1682 1896
proc. 3 proc. 3 proc. 3 proc. 1b proc. 1c proc. 2 and 3 proc. 3
GC/ MS/MS, SIM in CI GC/MS/MS, SIM in CI GC/MS, SIM in EI GC/MS, full scan in EI ret. index on DB-Wax and SPB-1 GC/MS-SIM in EI GC/MS/MS, SIM in EI
none none unclearc Schreier et al., 1976 Shiratsuchi et al., 1994 Badings and Neeter, 1980 none
a
Indices found by sniffing. b Procedure used for the sample preparation. c Bosset and Gauch, 1988.
Figure 3. GC/MS/MS chromatograms of the daughter ion 69 obtained by SRM from the [M - H]- ion at m/z 139: (lower trace) 50 pg of reference compound; (upper trace) 5 µL of yogurt extract.
Figure 2. Mass spectra of 1-nonen-3-one. (A) EI ionization spectrum. (B) Negative CI spectrum.
Impact Flavorings. A SNIF profile of the yogurt flavor was established using the headspace sampling procedure 1c, from 8 sniffing runs as explained above. Each headspace sampling was performed with a new yogurt originating from the same fermentation batch. Twenty main contributors to the overall aroma were detected (Figure 1 and Table 2). Among them, only 11 corresponded to volatiles that were identified using the three procedures mentioned in Table 1. On the basis of their GC and sensorial characteristics the identities
of seven other compounds were proposed by comparing retention indices and odor descriptors to those recorded in our sensorial data-base. Their retention indices were checked by re-injecting authentic samples under the same conditions using a FID detector. Identification proposals of the sensorial data base were confirmed by GC/MS and GC/MS/MS experiments (Table 3). Because of the low amount of each compound, full scan spectra could not be obtained by mass spectrometry in every case. Therefore, the compounds were selectively analyzed either by selected ion monitoring (SIM) of specific ions produced after electron impact or by collision-induced dissociation. Parent ions [M + NH4]+ or [M - H]- were generated after positive or negative chemical ionization, respectively. In this case,
Aroma Impact Compounds of Yogurt Flavor
J. Agric. Food Chem., Vol. 45, No. 3, 1997 855
Figure 4. NIF profile of a blank experiment (headspace cell loaded with distilled water, procedure 1c).
Figure 5. Determination of the 1-nonen-3-one threshold. C, concentration expressed in ng/kg.
the characteristic daughter ions obtained for each compound were recorded [selective reaction monitoring (SRM) experiment] if eluted from the GC column at the same retention time as the reference molecule. As an example, Figure 2A presents the electron impact ionization mass spectrum of 1-nonen-3-one and Figure 2B its mass spectrum after collision-induced dissociation of the [M - H]- ion. 1-Nonen-3-one was looked for in the yogurt extract after monitoring the daughter ion at m/z 69 from its parent ion at m/z 139. The GC/MS/MS traces obtained for a yogurt extract and for a standard of the reference compound are presented in Figure 3. Of the seven flavorings of Table 3, six of them [1-octen-3-one, 1-nonen-3-one, methional, 2-methyltetrahydrothiophen-3-one, (2E)-nonenal, and guaiacol] were detected for the first time in the yogurt flavor. They were probably missed in the previous works due to their low concentrations, although their low odor thresholds strongly influence the aroma profile. Three odor contributors remain unindentified (indices 1715, 1750, and 2043 of the aromagram).
A blank experiment was run by loading the headspace cell with distilled water instead of yogurt. Some peaks corresponding to those found in yogurt were detected, but with a lower intensity (Figure 4). As 3-methylbutyric acid and hexanoic acid were also found without using a Tenax trapping (sampling procedures 2 and 3), they do not correspond to artefacts. In addition, all peaks of the blank, except 1-nonen-3-one, have already been mentioned as occurring in milk and/or yogurt (Table 1). Consequently, odor contributors of Figure 1 originate from the yogurt flavor itself. Special attention was paid to 1-nonen-3-one, as no spectral evidence was ever given for this compound in any flavor. 1-Nonen-3-one: a New Impact Flavoring (Patent Pending; Chaintreau et al., 1996). To our knowledge, 1-nonen-3-one has not been positively identified in any food flavor (Maarse et al., 1994), and this compound is not commercially available. The only unambiguous occurrence was mentioned in the odorants produced by the thermal oxidation of polyethylene (Bravo et al., 1992). Even the corresponding alcohol, 1-nonen-3-ol, seems to be a non-common aroma component: only three natural occurrences are reported: chervil, beef fat, and banana (Maarse et al., 1994). A reference sample was then prepared to elucidate its sensorial and analytical properties. In addition to its mass spectra (Figure 2), its retention indices on different phases were determined (Table 4). According to quantity of the reference compound used in Figure 3, the 1-nonen-3-one concentration in yogurt seems to be in the 2-5 ppt range. As 1-nonen-3-one was also detected in the blank sniffing experiment, one could wonder whether the Tenax degradation could generate this compound like in the Bravo’s experiment of polyethylene degradation. As the positive spectral identification was performed on the SDE extract (procedure 3), this hypothesis can be excluded.
856 J. Agric. Food Chem., Vol. 45, No. 3, 1997
Ott et al.
Figure 6. NIF profile of milk aroma (isolated according to procedure 1c). Table 4. Retention Indices and Odor Description of 1-Nonen-3-one Occurring in Some Food Flavors sample
DB-Wax
SE-54 DB-5
FFAP
OV-1701
SPB1
1-nonen-3-one (authentic)
1400a
1080a
1398a
1171a
1070a
mushroom, earthy
yogurtc raspberryc troutc process flavorc toasted wheat breadc buttermilk off-flavorc
1425b 1424b 1405 1415 -
1065b 1065b 1052 -
mushroom, earthy mushroom mushroom, earthy mushroom mushroom mushroom and phenolic
1081 1077 1076
-
-
descriptor
a 1-Nonen-3-one injected as solution. b 1-Nonen-3-one absorbed on Tenax and injected with the ATD400 system. c See bibliographic references in the text.
Table 5. Most Potent Flavors and Off-Flavors
compound 2,3,6-trichloroanisole 2,3,6-tribromoanisole 1-nonen-3-one 2(5H)-furanone, 5-ethyl, 3-hydroxy, 4-methyl 2-isobutyl-3-methylpyrazine
odor threshold (pg/kg)
reference
0.3 8 8 10
Ewender et al., 1995 Ewender et al., 1995 this work Kobayashi, 1989
20
Ewender et al., 1995
Odor Threshold. The threshold was determined by sniffing a series of triangle tests of an aqueous solution, and panelists answers were treated according a reported method (Voirol and Daget, 1986). In a forced-choice answer procedure some correct answers are due to chance. Results were therefore corrected using the following formula:
% perception ) % correct answers (1 - % correct answers)/2 Perception percentages were transformed to probit values using the table established by Bliss (1967). Probit values were plotted against the logarithm of the
concentrations. The value for 50% perception (or for a probit value of 5) defined as perception threshold was calculated to be 8 pg/kg (Figure 5). This extremely low threshold classifies 1-nonen-3-one among the most potent compounds identified in flavors and off-flavors up to now (Table 5). Consequently, its concentration must be very low in any flavor, which explains why its identification was not spectrally proved up to now. Other Occurrences. In the literature, 1-nonen-3one can be suspected: among the main odorants revealed by aroma extract dilution, Milo, Rychlik, and Heiler reported an unknown compound in boiled trout, in toasted wheat bread, and in buttermilk off-flavor, respectively, eluting exactly 100 retention index units after 1-octen-3-one (Milo and Grosch, 1993; Rychlik and Grosch, 1996; Heiler and Schieberle, 1996). In all cases, its odor description and its index (Table 4) fit well with our finding. More recently, Blank suspected its occurrence in a process flavor (no retention index), with lipid degradation products such as (2E)-nonenal (Blank et al., 1994). As trout, yogurt, and buttermilk are fatcontaining foods as well, 1-nonen-3-one could also originate from unsaturated fatty acid oxidation. However, on the basis of its odor and retention indices, it was also
J. Agric. Food Chem., Vol. 45, No. 3, 1997 857
Aroma Impact Compounds of Yogurt Flavor Table 6. Comparison of Odor Impact Contributions in Milk and Yogurt Using SNIF Values ret. ind.a 716 757 995 1082 1120-1 1220-1 1322 1406 1424 1462 1479 1492 1522 1551
compound
1555 1624 1680-8
acetaldehyde dimethyl sulfide 2,3-butanedione 2,3-pentanedione 2-methylthiophene 3-methyl-2-butenal 1-octene-3-one dimethyl trisulfide 1-nonene-3-one acetic acid methional pyrazine (isomer?) pyrazine (isomer?) 2-methyltetrahydrothiophene-3-one and (2E)nonenal unknown pyrazine (isomer?) 2-phenylacetaldehyde
1680-4 1715-6 1750-3 1824 1882-7 1896-9 2002-12 2043
3-methylbutyric acid unknown unknown unknown hexanoic acid guaiacol benzothiazole unknown
milk 5269 3069 0 3487 2866 5282 3109 3114 3223 0 2734 3421 0
yogurt 7006 3969 8561 3549 2480 1804 2357 1804 1786 2051 2581 1007 0 2410 (both peaks)
1531 0 1951 0 3199 1113 (both peaks) 2339 1368 1344 1017 2507 1119 0 4470 4048 1315 3377 1779 3486 0 1125
2,3-pentanedione, and dimethyl sulfide are confirmed to be impact flavorings of yogurt. On the other hand, benzaldehyde does not seem to significantly participate to the overall aroma. Five compounds [1-nonen-3-one, methional, 2-methyltetrahydrothiophen-3-one, (2E)-nonenal, and guaiacol] with intense odor were found for the first time in the yogurt flavor. This paper also reports the first spectral identication of 1-nonen-3-one in a flavor. This component appears to be one of the most potent flavorings which has been identified up to now. Its low odor threshold justifies its impact role in yogurt in spite of its low concentration. This work also shows that the yogurt flavor results from the addition of aroma compounds originating from milk and of specific odor contributors generated by the metabolism of the lactic acid bacteria. ACKNOWLEDGMENT
We gratefully acknowledge Prof. Couturier (University of Lille, France) for the synthesis of authentic 1-nonen-3-one, Mr. Y. Krebs for MS spectra, Dr. Voirol and D. Cazes for their advice in sensorial measurements, Dr. D. Roberts for English revision, and all colleagues who participated in the sniffing panels. LITERATURE CITED
a
Retention index on DB-Wax phase, using headspace injection (procedure 1c).
detected in raspberry, a non-fat product (Roberts and Acree, 1996). Origin of Yogurt Impact Flavorings. With the aim of identifying flavorings originating from the strain action, the aromagram of the milk mixture (skimmed milk powder and pasteurized full fat milk) was established (Figure 6). The milk amount was the same as used for the yogurt preparation. As frequencies of detection of a given peak (Pollien et al., in press) increased with concentration, yogurt and milk aromagrams were compared using SNIF values as the areas take into account possible minor variations of retention for each of the eight assessors. Many peaks were common to both aromas: impact flavorings already present in milk exhibited similar or lower SNIF values in yogurt (Table 6). On the other hand, SNIF values of several compounds [acetaldehyde, 2,3-butanedione, 2,3pentanedione, methional, 2-methyltetrahydrothiophene3-one, and (2E)-nonenal], were greatly increased by the fermentation step. Levels of dimethyl trisulfide, guaiacol, and benzothiazole could be also affected, and this should be verified by a quantitative analysis. Pyrazines (isomers not determined) probably occurred in milk due to its thermal treatment, but they were not observed to contribute any more to the yogurt aroma. Consequently, yogurt flavor appears to result from milk odor contributors and specific flavorings synthesized by the lactic acid bacteria as SNIF differences were equal or less than the least significant difference for this method (Pollien et al., in press). CONCLUSION
Using the new SNIF approach for the data treatment of GC-olfactometric signals, the main contributors of the yogurt odor have been determined. Of the 91 volatiles identified in the flavor, 21 exhibit a key aroma impact. According to Imhof (1995), 2,3-butanedione,
Abbott, N.; Etievant, P.; Issanchou, S.; Langlois, D.; Critical evaluation of two commonly used techniques for the treatment of data from extract dilution sniffing analysis. J. Agric. Food Chem. 1993, 41, 1698-1703. Acree, T. E.; Barnard, J.; Cunningham, D. G. The analysis of odor-active volatiles in gas chromatographic effluents. In Analysis of Volatiles; de Gruyter: New York, 1984. Badings, H. T.; Neeter, R. Recent advances in the study of aroma compounds of milk and dairy products. Neth. Milk Dairy J. 1980, 34, 9-34. Bemelmans, J. M. H. Progress in Flavour Research; Land, D. G., Nursten, H. E., Eds.; Applied Science: London, 1979, pp 79-98. Blank, I.; Sta¨mpfli, A.; Eisenreich, W. Trends in Flavor Research; Maarse, H., van der Heij, D. G., Eds.; Elsevier Sciences B.V.: Amsterdam, 1994. Bliss, C. I. Statistics in Biology; McGraw-Hill: London, 1967. Bosset, J. O.; Gauch, R. Effet protecteur de l’emballage contre la photooxydation (Protective effect of packaging against photooxidation). Trav. Chim. Aliment. Hyg. 1988, 79, 165174. Bravo, A.; Hotchkiss J. H.; Acree, T. E. Identification of odoractive compounds resulting from thermal oxidation of polyethylene. J. Agric. Food Chem. 1992, 40, 1881-1885. Buttery, R.; Ling, L. C.; Guadagni, D. G. Volatilities of aldehydes, ketones, and esters in dilute water solution. J. Agric. Food Chem. 1969, 17, 385-389. Chaintreau, A.; Grade, A.; Mun˜oz-Box, R. Determination of partition coefficients and quantitation of headspace volatile compounds. Anal. Chem. 1995, 67, 3300-3304. Chaintreau, A.; Maignial, L.; Pollien, Ph. Appareil et proce´de´ de distillation et d’extraction simultane´es (Apparatus and process for simultaneous distillation and extraction). Eur. Pat. 95 203 465.5, 1995. Chaintreau, A.; Fay, L. B.; Ott, A. Pat. EP 96202184.6, 1996. Corey, E. J.; Suggs, J. W. Pyridinium chlorochromate. An efficient reagent for oxidation of primary and secondary alcohols to carbonyl compounds. Tetrahedron Lett. 1975, 31, 2647-2650. Ewender, J.; Lindner-Steinert, A.; Ru¨ter, M.; Piringer, O. Sensory problems caused by food packaging interactions: overview and treatment of recent case studies. In Foods and
858 J. Agric. Food Chem., Vol. 45, No. 3, 1997 packaging materials: Ackermann, P., Ja¨gerstad, M., Ohlsson, T., Eds.; Royal Society of Chemistry: Cambridge, 1995, 33-44. Go¨rner, V.; Palo, V.; Bertan, M. Vera¨nderungen des Gehaltes der flu¨chtigen Stoffe wa¨hrend der Jogurtreifung (Changes in the content of volatiles during ripening of yogurt). Milchwissenschaft 1968, 23, 94-100. Heiler, C.; Schieberle, P. Studies on the metallic off-flavour in buttermilk: identification of potent aroma compounds. Lebensm.-Wiss. -Technol. 1996, 29, 460-464. Hild, J. Quantitative Bestimmung einiger wichtiger Aromakomponenten in Milcherzeugnissen mittels Head-SpaceTechnik (Quantitative determination of some impact components in dairy product by a headspace technique). Milchwissenschaft 1979, 34, 281-283. Imhof, R.; Bosset, J. O. Quantitative GC-MS analysis of volatile flavour compounds in pasteurized milk and fermented milk products appplying standard addition method. Lebensm.-Wiss. -Technol. 1994, 27, 265-269. Imhof, R.; Gla¨ttli, H.; Bosset, J. O. Volatile organic aroma compounds produced by thermophilic and mesophilic mixed strain dairy starter cultures. Lebensm.-Wiss. -Technol. 1994a, 27, 442-449. Imhof, R.; Gauch, R.; Sieber, R.; Bosset, J. O. U ¨ ber einige flu¨chtige organische Verbindungen in Milch und Milchprodukten Mitt (Some volatile organic compounds in milk and milk products). Gebiete Lebensm. Hyg. 1994b, 85, 681703. Imhof, R.; Gla¨tti, H.; Bosset, J. O. Volatile organic compounds produced by thermophilic and mesophilic single strain dairy starter cultures. Lebensm.-Wiss. -Technol. 1995, 28, 78-86. Kang, Y.; Frank, J. F.; Lilliard, D. A. Gas chromatographic determination of yogurt flavor compounds and changes during refrigerated storage. Cultured Dairy Prod. J. 1988, 11, 6-9. Kobayashi, A. Sotolonsidentification, formation and effect on flavor. In Flavor chemistry: trends and development; Teranishi, R., Buttery, R. G., Shahidi, F., Eds.; ACS Symposium Series 388; American Chemical Society: Washington DC, 1989, pp 49-59. Laye, I.; Karleskind, D.; Morr, V. V. Chemical, microbiological and sensory properties of plain nonfat yogurt. J. Food Sci. 1993, 58, 991-995. Maier, I.; Fieber, M. Retention characteristics of volatile compounds on Tenax TA. J. High Resolut. Chromatogr. 1988, 11, 566-579. Maarse, H.; Visscher C. A. Volatile compounds in food, TNOCIVO; Food Analysis Institute: The Nederlands, 1994. Maignial, L.; Pibarot, P.; Bonetti, G.; Chaintreau, A.; Marion, J. P. Simultaneous distillation-extraction under static vacuum: Isolation of volatile compounds at room temperature. J. Chromatogr. 1992, 606, 87-94. Milo, Ch.; Grosch, W. Changes in the odorants of boiled trout (Salmo fario) as affected by the storage of the raw material. J. Agric. Food Chem. 1993, 41, 2076-2081.
Ott et al. Nakahira, H.; Ryu, I.; Ikebe, M.; Oku, Y.; Ogawa, A.; Kambe, N.; Sonoda N.; Murai, S. β-Trichlorostannyl ketones and aldehydes. Preparation and facile amine-induced dehydrostannation leading to R-methylene ketones and aldehydes. J. Org. Chem. 1992, 57, 17-28. Ott, A.; Pollien, P.; Montigon, F.; Chaintreau, A. Improved headspace-GC-sniffing technique: screening of aroma impact flavorings. 4th International Symposium on Hyphenated Techniques in Chromatography, February 7-9, 1996, Bruges, Belgium, 1996. Pollien, P.; Ott, A.; Montigon, F.; Baumgartner, M.; Mun˜ozBox, R.; Chaintreau, A. Hyphenated headspace-GC-sniffing technique: screening of impact odorants and quantitative aromagram comparisons. J. Agric. Food Chem. (in press). Roberts, D.; Acree, T. Effects of heating and cream addition on fresh raspberry aroma using a retronasal aroma simulator and gas chromatography olfactometry. J. Agric. Food Chem. 1996, 44, 3319-3925. Rychlik, M.; Grosch, W. Identification and quantification of potent odorants formed by toasting of wheat bread. Lebensm.-Wiss. -Technol. 1996, 29, 515-525. Schreier, P.; Drawert, F.; Junker, A.; Barton, H.; Leupold, G. Biosynthesis of aroma compounds by microorganisms. II. Formation of sulfur compounds from methionine by Saccharomyces cerevisiae. Z. Lebensm. Unters. Forsch. 1976, 162, 279-284. Semmelroch, P.; Grosch, W. Analysis of roasted coffee powders and brews by gas chromatography-olfactometry of headspace samples. Lebensm.-Wiss. - Technol. 1995, 28, 310313. Shiratsuchi, H.; Shimoda, M.; Imayoshi, K.; Noda, K.; Osajima, Y. Volatile flavor compounds in spray-dried skim milk. J. Agric. Food Chem. 1994, 42, 984-988. Ulberth, F. Headspace gas chromatographic estimation of some yogurt volatiles. J. Assoc. Off. Anal. Chem. 1991, 74, 630634. Ullrich, F.; Grosch,W. Identification of most intense volatile flavor compounds formed during autooxidation of linoleic acid. Z. Lebensm. Unters. Forsch. 1987, 184, 277-282. Van den Dool, H.; Kratz, P. D. A generalization of the retention index system including linear temperature programmed gas-liquid partition chromatography. J. Chromatogr. 1963, 11, 463-471. Voirol, E. ; Daget, N. Comparative study of nasal and retronasal olfactory perception. Lebensm.-Wiss. -Technol. 1986, 19, 316-319. Received for review July 8, 1996. Revised manuscript received November 25, 1996. Accepted November 26, 1996.X JF960508E X Abstract published in Advance ACS Abstracts, February 1, 1997.
